Improved automatic identification of isolated rapid eye movement sleep behavior disorder with a 3D time‐of‐flight camera

Abstract Background and purpose Automatic 3D video analysis of the lower body during rapid eye movement (REM) sleep has been recently proposed as a novel tool for identifying people with isolated REM sleep behavior disorder (iRBD), but, so far, it has not been validated on unseen subjects. This study aims at validating this technology in a large cohort and at improving its performances by also including an analysis of movements in the head, hands and upper body. Methods Fifty‐three people with iRBD and 128 people without RBD (of whom 89 had sleep disorders considered RBD differential diagnoses) were included in the study. An automatic algorithm identified movements from 3D videos during REM sleep in four regions of interest (ROIs): head, hands, upper body and lower body. The movements were divided into categories according to duration: short (0.1–2 s), medium (2–15 s) and long (15–300 s). For each ROI and duration range, features were obtained from the identified movements. Logistic regression models using as predictors the features from one single ROI or a combination of ROIs were trained and tested in a 10‐runs 10‐fold cross‐validation scheme on the task of differentiating people with iRBD from people without RBD. Results The best differentiation was achieved using short movements in all four ROIs (test accuracy 0.866 ± 0.007, test F1 score = 0.783 ± 0.010). Single group analyses showed that people with iRBD were distinguished successfully from subjects with RBD differential diagnoses. Conclusions Automatic 3D video analysis might be implemented in clinical routine as a supportive screening tool for identifying people with RBD.


Figure S1 :
Figure S1: Correlation between the SINBAR RWA index and test probability of iRBD

Table S2 : Demographic and sleep information of iRBD and patients with restless legs syndrome.
Values are shown as mean±standard deviation if normally distributed and as median [interquartile range] otherwise.For normally distributed variables, t-tests were used to compare groups, otherwise Mann-U-Whitney tests were employed.Categorical variables were compared with chi-square tests.Significant p-values (<0.05) are highlighted in bold.

Table S3 : Demographic and sleep information of iRBD and patients with PLMS without any associated sleep disorder.
Values are shown as mean±standard deviation if normally distributed and as median [interquartile range] otherwise.For normally distributed variables, t-tests were used to compare groups, otherwise Mann-U-Whitney tests were employed.

Table S5 : Demographic and sleep information of iRBD and patients with non-REM parasomnia. Values
Legend: AHI: apnea hypopnea index; iRBD: isolated REM sleep behavior disorder; PAP: positive air pressure; PLMS: periodic limb movement during sleep; REM: rapid eye movement; SINBAR: sleep Innsbruck Barcelona.

Table S6 : Demographic and sleep information of iRBD and subjects without any relevant sleep disorder
. Values are shown as mean±standard deviation if normally distributed and as median [interquartile range] otherwise.For normally distributed variables, t-tests were used to compare groups, otherwise Mann-U-Whitney tests were employed.Categorical variables were compared with chi-square tests.Significant p-values (<0.05) are highlighted in bold.Legend: AHI: apnea hypopnea index; iRBD: isolated REM sleep behavior disorder; PAP: positive air pressure; PLMS: periodic limb movement during sleep; REM: rapid eye movement; SINBAR: sleep Innsbruck Barcelona.

Table S7 : Performances in the training folds for the classification iRBD vs no-RBD.
The performance measures are reported for each interval duration and when considering as predictor features the 3D rate and 3D ratio from: i) head region of interest (ROI) only (HE), ii) Legend: NPV: negative predictive value; PPV: positive predictive value.

Table S9 : P-values from the comparison of performances considering different movement lengths for the classification iRBD vs no-RBD.
The table reports the corrected p-values obtained from comparing test accuracy and F1-scores of classifiers trained and

Table S10 : P-values from the comparison of performances considering different ROIs and short movements (classification iRBD vs no-RBD).
The table reports the corrected pvalues obtained from comparing test accuracy and F1-scores of classifiers trained and tested using short movements when considering as predictor features the 3D rate and 3D ratio from:

Table S12 : P-values from the comparison of performances considering different groups.
The table reports the corrected p-values obtained from comparing test accuracy and F1-scores of classifiers trained and tested using short movements (0.1s-2s) and for different classification problems: i) iRBD vs SRBD, ii) iRBD vs RLS/PLMS and iii) iRBD vs INS/NRSD.Legend: INS: insomnia; iRBD: isolated rapid eye movement sleep behavior disorder; NRSD: non-relevant sleep disorder; PLMS: periodic limb movements during sleep; RLS: restless legs syndrome; SRBD: sleep-related breathing disorder.

Table S13 : Details of the regression model evaluating the influence of age, sex and groups on the classification
. The test probability of the 10 runs of classification considering short movements (0.1s-2s) in the four ROIs were averaged to obtain the values of p(iRBD).The following linear regression model was fit: log(p(iRBD)) ~ 1+age+sex + iRBD + SRBD + RLS + PLMS + Insomnia + NREM parasomnia.The sex and groups were considered categorical variables.The log-transformation for p(iRBD) was done to ensure normal residuals.